Loading scores...
Player Analysis

Don't Believe the Hype

by Galin Dragiev
Updated On: October 4, 2018, 4:09 pm ET


Anytime we have 10 days off without any Premier League action, it allows me to take a step back and look at the mistakes I’ve made in the past two weeks.  My biggest mistake of week 26 was including Kevin Nolan in my squad.  Coming off of two weeks with at least 24 points and facing a poor Norwich side at home, what could have possibly gone wrong?  If you didn’t see, he scored 1(yes, one) point.  Another big mistake was playing Christian Eriksen in week 23, when he was coming off an incredible 4-5 game run, but he ended up giving me a poor 4 points.


I was wondering where I had gone wrong with those two decisions so I decided to check how wise it really is to choose players who are on a short run of good form.  To do that, I created three statistical models for predicting a player’s performance.  The first model uses the player’s form from the past three games, the second uses his “point per game(PPG)” average from the entire season, and the third combines both form and PPG.  I ran some numbers to check which model predicts a players’ score the best.


The numbers

I found the error that each model produced for every player for every week, and then saw which model produced the most error.  The error is simply the difference between the predicted score for each model and the actual score of the player for that week.  I then took the mean and standard deviation of the scores for each model.  The smaller the mean, the less difference there was between the predicted value from the model and the actual value of the players’ score.  In other words, a low mean is a better model.  A smaller standard deviation is also representative of a better model.


I’d like to thank Professor M over at Never Manage Alone for providing me with the data used for this task.




Standard Deviation

Form (last 3 games)



Points per game (season)



Combo (form and PPG)





What it all means

1.  Form doesn’t usually last long: We see this way too often.  A player has two, three, maybe four great weeks and everybody jumps on the bandwagon. Kevin Nolan followed up 32 and 24 points with a 1 point outing.  Christian Eriksen averaged 16.4 points over five weeks before getting back to earth to average 3.33 points over his next three games.  It’s just not very wise to rely on form as a way to predict how well a player will do on a given week.


If you order Liverpool’s players by form right now, you’ll find Daniel Sturridge and Raheem Sterling in first place with 15 each.  But does that really mean they’re equivalent picks for next week?  The numbers say no, so what do we look at instead?


2.  Season averages:  The better choice is to look at players who have been doing well all season.  The average error decreased by more than 11% when going from the form model to the PPG model, which can be fairly significant over the course of a season.  This can also help us in identifying bargains.  Players like Aaron Ramsey and Daniel Sturridge, who have both missed significant time due to injuries, are probably good to pick up upon their return.  Taking advantage of cheap prices can be a game changer, and as we’ve seen with Sturridge, many are as good post injury as they were before it.  Someone else that fits the bill at moment is Ross Barkley, who is in terrible form but has been a decent performer on the year.  I wouldn’t be surprised to see him score well after he returns from injury.


3.  The interesting thing was that the third model, a combination of form and PPG, didn’t really improve the predictions.  The result is probably due to too much noise in the form, as one great game (as in Sterling’s case), has too big of an impact.  I’ll examine the issue again at the end of season when I have twice the amount of data, but for now, I’d rather have a proven performer over someone who is on a good run of two games.


More than just numbers

I was talking about this to some of the guys here at rotoworld and the biggest questions coming out of this discussion was:  When is “good form” no longer just “form”, but a result of a young player improving, a shift in teams, or a managerial change?  Unfortunately, that question is much tougher to answer.  I could try to answer it mathematically, and I probably will in the near future, but every situation is too different.  Coaching changes can impact players in ways we could have never imagined.  A young player gaining confidence can result in the instant creation of a star.  A shift in teams can give players new opportunities in ways not even their coaches could have imagined.  The options are endless.  And at the end of the day, that’s the best part about sports:  a combination of hope, extremely hard work, and a great deal of luck.

Galin Dragiev
Galin Dragiev is a Rotoworld contributor that focuses on Daily Fantasy Soccer and contributes for a number of outlets in the industry. You can find him on Twitter at @GalinDragiev.